London Fashion Week finished this past week (strangely, I wasn’t invited) so with my sometime obsession around interesting applications of data streams, it’s timely to think about what the collision of two such polar opposite worlds as fashion and data might look like. Fortunately there’s already some interesting examples.
Take Fashiolista. Backed by one of the founders of Skype, Fashiolista provides a “Love” button (rather similar to a Facebook “Like” button) through a user installed toolbar that allows you to “love” items as you browse shopping sites, build your own collection, and follow other people and collections. Apparently, over 1 million items are currently “loved” per month. But the real potential here of-course is data driven – being able to build up profiles of people and their preferences in order to power recommendations and facilitate connection with likeminded people. In other words scrobbling fashion.
And then we have Google’s Boutiques.com, a site that is designed to not only make it easier to find clothes you like, but to actually predict what those clothes will be. As well as a series of ‘Boutiques’ curated by designers, fashion bloggers and celebrities which you can browse, the site uses a Hunch-like quiz designed to pick out quirks of personality and taste, which then decide which style bucket you fall into. Visual search technology is used to help create a personalised Boutique, make recommendations, and in theory get you to the stuff you like quicker.
But the ultimate use of data of-course is to be able to predict fashion trends. Which is just what EditD (“fashion industry intelligence in real-time”) is trying to do, generating 18 month fashion forecasts by analysing streams of internet data from blogs, status updates and tweets (up to 600,000 a day apparently) using natural-language processing to measure sentiment, and attributing a measure of influence to each datapoint. Geoff Watts, computer scientist and co-founder, points out that fashion forecasting previously worked on little more than educated hunches: ” Those hunches worked well when the industry was based on six-month catwalk cycles, but fashion nowadays is moving too fast. There are no ‘seasons’ any more.”
There’s been some really interesting examples of community-driven innovation in fashion recently (Uniqlo with their Burberry Art of the Trench-like Uniqlooks, ASOS opening their facebook store, the NY Times curating readers photos of NY fashion week, and models such as GoTryItOn) but fashion and data is an equally fascinating mix. The real challenge of-course for all data-driven models in this space is getting the balance right between the art and the science.
P.S. the guys at MadeByMany have put together a really useful list of fashion start-ups
P.P.S. In case you’re wondering, I take my own fashion inspiration from the guy on the left.
Original Post: http://neilperkin.typepad.com/only_dead_fish/2011/02/fashionable-data.html
